Addressing the issues of low efficiency in manual adjustment of process parameters (such as powder removal volume and separation density) in the de-powdering screening and dense medium separation stages of coal preparation plants, which lead to significant fluctuations in product quality, this study proposes an intelligent decision-making and collaborative control method for the coal separation process based on the multi-objective Whale Optimization Algorithm (WOA). The method innovatively extends WOA into a multivariable optimization framework, constructing a composite fitness function to balance ash content deviation, clean coal yield, and process constraints, thereby achieving automated parameter adjustment. Using a coal preparation plant in Jincheng, Shanxi, as a case study, an ash content-yield prediction model is established by integrating historical production data and sink-float tests. Experimental results demonstrate that, compared to manual adjustment, WOA reduces the average ash content deviation (the average deviation between actual and target ash content) from 1.5% to 0.85%, increases the clean coal yield from 38.82% to 41.73%, and improves the calorific value qualification rate from 56.0% to 85.0%, significantly reducing quality surplus and nearly eliminating unqualified products. In comparison with Particle Swarm Optimization (PSO), WOA shows superior performance in terms of ash content deviation, clean coal yield, and calorific value qualification rate; at the same time,validation using subsequent production data confirms its generalization ability. This method provides theoretical support and practical reference for the intelligent optimization of coal preparation processes, demonstrating the effectiveness and robustness of WOA in addressing multivariable, nonlinear industrial problems.